Scalable hybrid reliability frameworks blending edge computing and cloud AI for real-time equipment health monitoring in the mining sector.
Mining operations face a fundamental contradiction in reliability engineering: the assets that most need continuous monitoring — draglines, thickeners, ball mills, conveyors — are often located in environments with limited connectivity, harsh conditions, and massive data volumes that overwhelm traditional cloud-only architectures.
Existing reliability frameworks treat edge and cloud as separate choices. My research argues they must be integrated — with intelligent partitioning of processing between edge devices at the asset and cloud platforms for fleet-wide pattern recognition, all grounded in rigorous RCM2 methodology rather than pure data-driven approaches.
The goal is a scalable hybrid reliability framework that works in the real conditions of Australian mining — intermittent connectivity, diverse equipment types, regulatory requirements, and the practical constraints of maintenance teams in the field.
Develop an integrated reliability model that works across heterogeneous equipment types — from rotating machinery to process instrumentation — within a single unified framework.
Design and validate a hybrid edge-cloud infrastructure that partitions processing intelligently — real-time anomaly detection at the edge, fleet-wide pattern learning in the cloud.
Validate digital twin deployment on edge and hybrid systems, establishing performance benchmarks for accuracy, latency, and resource consumption in mining environments.
Integrate RCM2 methodology as the analytical backbone of the framework — ensuring AI-generated maintenance recommendations are grounded in engineering consequence logic, not just statistical pattern matching.
Develop and validate algorithms for multi-equipment model generalisation — so the framework learns from failures across the fleet and transfers that knowledge to individual assets.
Conduct comparative validation with real-world industrial partners, comparing framework performance against existing maintenance approaches on production mining equipment.
Comprehensive review of existing reliability frameworks, edge computing architectures, digital twin methodologies, and RCM2 applications in mining. Gaps identified in hybrid edge-cloud integration.
Established simulation environment for initial model testing using synthetic and public datasets. Baseline performance metrics established for comparison with proposed framework.
Published preliminary findings at international conference on digital twins. Framework architecture presented and validated by peer review. Feedback incorporated into refined research design.
Optimising the algorithm for transferring reliability models across heterogeneous equipment types. Current focus on thickener, conveyor, and rotating machinery datasets.
Deploying prototype hybrid infrastructure on test environment. Edge nodes running lightweight anomaly detection; cloud layer performing fleet-wide pattern analysis and RCM2 decision logic.
Comparative validation with industrial partners on production mining equipment. Framework performance benchmarked against existing maintenance approaches across multiple asset types.
Final thesis documenting the complete hybrid reliability framework, validation results, and contribution to reliability engineering literature.
The hybrid framework operates across four integrated layers — from physical assets at the edge to fleet-wide intelligence in the cloud.
Physical mining equipment with embedded sensors — vibration, temperature, pressure, current. Raw condition data generated continuously in the field.
On-site edge devices running lightweight ML models for real-time anomaly detection. Low latency, operates on limited connectivity, generates alerts without cloud dependency.
Fleet-wide pattern recognition, digital twin synchronisation, RCM2 decision logic, and long-term reliability trend analysis. Learns from failure events across all assets.
AI-powered maintenance recommendations grounded in RCM2 methodology. Consequence-first logic ensures recommendations are defensible engineering decisions, not statistical outputs.
Research partnerships, industrial validation opportunities, or discussions on hybrid reliability frameworks — get in touch.